You are here

tools to transcribe audio to text?

9 posts / 0 new
Last post
euxalot
tools to transcribe audio to text?

Love the audio collections here BUT text is still king when you want to search and datamine specific terms and topics. It's also much faster to just scan ahead if anything looks interesting.

Can anyone recommended tools to transcribe audio to text? I see a lot of paid tools out there but have not evaluated them.

would welcome thoughts and suggestions.

Best practice recommendation when uploading:

Keep audio files *with* the accompanying text; and
Keep PDFs paired with epubs since you can search across a folder of PDFs but not epubs

euxalot
Possibilities include:

Possibilities include:

1. Enable voice typing on your word-processing software. In Google Docs go to “Tools -> Voice typing” to turn on dictation. For Microsoft Word you need to turn on dictation/online speech recognition on macOS or Windows 10, then press the keyboard shortcut you set to turn on dictation (Win + H in Windows 10, Command key twice in Mac) and open Word. Once you’ve set up dictation or voice typing, open your word-processing software, then play the audio file into your PC microphone (or just play it on your PC so your mic picks it up).

2. Youtube: since youtube has good captioning ability (the same thing as the paid services above) you can upload the file to youtube and then download it again but this time capturing the captions.

3. Android "Speechnotes"
https://play.google.com/store/apps/details?id=co.speechnotes.speechnotes
If your audio source is on a different device, you can use standard speech-to-text apps on your phone to transcribe the audio. Speechnotes is a highly rated Android app which does a pretty decent transcription. This method is absolutely free and gives nearly accurate results for audio files. However, you can only take notes separately and can’t transcribe the results into the audio or video source. (I didn't understand this final restriction so will update after I run some tests.)

4. Windows only "MP3 Speech to Text Converter" for 23 EU.
https://www.sobolsoft.com/convertmp3text/

conspyre
I've been using a paid service, it's not awful

Going with YouTube is a good idea, the price is right. I couldn’t because much of the media I want to do is verboten on that communist platform.

I went with web based, just comparison shopped what I found on a quick search and came up with https://otter.ai

For free, you can transcribe 600 minutes/month but are limited to 3 uploads and 40 minutes per file so, meh.

For $12.99/mo you can do 6000 minutes/month with unlimited uploads.

The higher tier adds collaboration features, minutes cap out at 6000.

The AI is pretty decent, about 98-99%, which means you have to review the entire transcript and make several edits per paragraph. To its credit, it has really good tools for editing such as skipping playback to the word you click on. It’s got some room for improvement in the accuracy department but its serviceable - you can transcribe somewhat faster than realtime.

It does a decent job of picking up audio over background, even music. The paid version also knows how to emit subtitle files.

euxalot
many thanks for sharing! that

many thanks for sharing! that is really helpful tips for all of us, methinks,
cheers

laneigile
Nuance Dragon
euxalot
cheers, Laneigile, it's

cheers, Laneigile, it's helpful

PS yep, my answers are indeed in Indian philosophy.
I think also in DMT.
And real trance work

anyone here try Indian Soma?

lifeispain
Anything on Linux to read pdf's would be great

Foxit Reader on Linux isn't like Foxit Phantom PDF Reader for Windows (cracked), which read me books and had a ton of different voices which sounded pretty realistic. The only place I use for this right now is inforobo, I don't trust the site as I have to do pkill -f speech after enabling bootstrap.cdn in noscript, it will make my speakers buzz very loudly when reading the text until I do that. Then it won't work, I have to refresh it and then defit.org shows up in noscript and I have to enable that and then it works, but only for Female UK and Male UK. :shrug:.

Also, in linux there is an easy tool to transform non-OCR pdf's (where one can't select text) into OCR pdf's, so I have a lot more than I could throw in such a software. Problem is since Win7 died, I'm not using any Windows anymore, at least not on my main desktop, the hi-fi entertainment centre in my livingroom which serves as a gaming console, home cinema, which is pretty unlimited considering the private trackers where my server paid off insanely over the years, I can download new stuff at any time, so that one can have WIn10, but the only internet it's doing is socks5 proxied to an sftp server that is my vps. Things of importance I keep to my main desktop which is now only Linux and always just be some distro of linux.

ConCen
lifeispain wrote:
lifeispain wrote:

Things of importance I keep to my main desktop which is now only Linux and always just be some distro of linux.

Good to see another Linux user here!

 

funkytunie
otter.ai and Telegram

I've tried otter.ai and honestly wasn't that impressed with the accuracy. Now for subscribed Telegram users, their voice message translator works extremely well from what I have heard & seen. Unfortunately, it is for voice messages recorded in app only as far as I know. Perhaps there is a workaround? Is anyone familiar with any newer and more accurate tools, like was said above there is no substitute for being able to search mass amounts of text in an instant. Thanks.

Log in to post comments